The presence of 5S variants in the same genome has been described in several mammals, amphibians and fishes [6, 25, 26] among others, and this suggests that variant 5S types originated early in the evolution of the vertebrates . Two types of 5S rDNA were also described in the Leporinus genus, and each was localized in a different locus, suggesting a separated evolution of the two types . However, the presence of two 5S rDNA variants is common in fishes [29–31], and this feature has been attributed to the ancient duplication event that occurred before the divergence of the main groups of teleost fishes . The presence of two variants in fishes has often been associated with a dual system where one of these variant is expressed in both somatic and oocyte cells, and the other variant only in oocyte cells, providing an increased physiological plasticity .
In the present work, up to four 5S rDNA variants were obtained (named from type α to type δ), which were characterized by their different NTS. The high evolutionary distance observed in the coding region of β and δ types leads to the hypothesis that they could be pseudogenes. The secondary structures of all 5S rRNA types (except β type and one clone of the δ type) showed secondary structures according to the models described . The abnormal structure observed in the other clone of the type δ at the Helix III-Loop C region might be deleterious, since mutations occurring in this region could have lethal effects or affect the translation accuracy ; therefore it is possible that the β and δ types of 5S coding region could be pseudogenes. Similarly, the clone of the γ type with a deletion which affects the Loop C size could also be a non-functional gene.
The birth-and-death model has been used to explain the evolutionary mechanism of the 5S rDNA in several organisms such a fungal species , razor clam species , and fish species . In the crustacean genus Pollicipes, up to seven different types of 5S rDNA and two pseudogenes have been described, suggesting a birth-and-death evolution pattern . Initially, the birth-and-death model seems to drive the evolution of the 5S rDNA in P. mediterraneus, because variants and pseudogenes have been seen. However, is this the only model which acts in the 5S evolution? As mentioned above, the β and δ types showed a striking homology with 5S rDNA from R. asterias, in both the coding and NTS regions. P. mediterraneus and R. asterias are fishes which belong to different classes (Actinopterygii and Chondrichthyes respectively), which separated 527 My ago . Two hypotheses can be outlined in this case: first, the β and δ types could have originated before the separation of the two classes, and could have been maintained in some lineages and lost in others. Nevertheless, the NTS are considered very dynamic regions of the genome, since they are free to mutate, and the variants which arise are almost neutral to natural selection; and they can be either fixed or lost, thus causing differences between closely related species, and even within individuals . The pseudogenization of the β and δ types would accelerate the accumulation of mutations in the NTS region, making this first hypothesis unlikely. A second hypothesis should not be discounted: a Horizontal Transference (HT) event. After HT occurs, the transferred gene either maintains its functionality and stays in the host genome, or loses its functionality and becomes a pseudogene. The β and δ types of the P. mediterraneus 5S rDNA seem to be in this second situation. Therefore, before the HT happened, P. mediterraneus had two types of 5S rDNA (types α and γ), as described for many fish orders . A process of duplication and specialization could have originated the types α and γ, as the birth-and-death model predicts, but within each type there is considerable homogeneity, as alignment within each type demonstrates. Hence, a mixed process could have driven the evolution of the 5S rDNA in P. mediterraneus, with the birth-an-death model acting at genome level and concerted evolution acting at locus level. This dual process has already been proposed for other fish species , and for the mollusc Mytilus spp. .
The precise mechanism by which the HT occurs is not yet well-understood, but there are two plausible mechanisms. One of these is mediated by transposable elements (TE) (see  for a review). The 5S rDNA family is an excellent candidate for this kind of HT, since interactions between retro-transposons and 5S rDNA have been demonstrated [40, 41], and this interaction could be “the door” for a 5S rRNA gene lateral transfer. Another possible mechanism should be considered: sperm-mediated gene transference (SMGT), since it is well-known that sperm cells are able to capture exogenous DNA and to transfer it to the oocyte at fertilization . The exogenous DNA is captured from the water column and the sediments, in which exogenous DNA is known to accumulate . This, taken together with the external fecundation of the fish species, makes this HT mechanism a possibility. An HT event could have great evolutionary consequences as a source of biological innovation; indeed it has been postulated that the origin of the primordial eukaryote cell or some characteristics of multicellularity have developed from HT events .
The type α presented its own peculiarities, like the NTS fragment widely represented in the NTS of other fish species. In this case, up to 11 species from several fish families share 27 bp of this region (see Figure 3). This sequence could have some regulatory function as the enhancer element like that of the mammals, named the D-box, which can increase the transcription up to 10 times . Also found within the same NTS fragment was an A-rich region (5’-CAAACAG-3’) similar to that found by  in some fish species (5’-GAAACAA-3’); those authors proposed this region as a second terminator region. The α type NTS also contains downstream another region very similar to the tRNA-Gln gene of various species. However, the P. mediterraneus secondary structure obtained with the RNAstructure program was different with respect to previously described models ; this, together with the higher accumulation of polymorphic sites in the P. mediterraneus tRNA sequence, leads to the proposal of the tRNA-Gln gene as a non-functional gene.
The 5S rDNA is known to be a multigene family which can be found linked with a variety of other multigene families, such as the major ribosomal genes, the histone genes, the trans-spliced leader genes, and the small nuclear RNA genes [9, 46, 47]; this indicates that the 5S rDNA has the ability “to jump” to other loci. This is not the first evidence of a 5S rDNA-tRNA linkage, since this kind of association has been found in a fungus species  and in various mussel species . Here we describe for first time this kind of linkage in a fish species.
The tRNA presented here could be a short interspersed element (SINE). The majority of SINEs derive mainly from tRNA sequences  and share some common features: i) a tRNA-like region with slightly modified A box and B box regions; ii) a tRNA-unrelated region; iii) an AT-rich region ending with poly-A [50, 51]. All these features are present in the type α NTS of P. mediterraneus, so it is possible that a tRNA-derived SINE is linked with the 5S rDNA type α. From an evolutionary point of view, it has been postulated that SINE integration into new localizations has potential implications, since it can disturb gene expressions or serve as a source of genomic innovation and a factor of genome plasticity .
The chromosomal localization of the 5S rDNA probe observed in P. mediterraneus (internal and near the centromere) is the most common pattern observed within fish species, so this could be the optimal localization for the arrangement of this gene . The 5S rDNA localization has been determined in six species of the Haemulidae family, and between two and four chromosomes showed positive signals [54–56]. Therefore, in the Haemulidae family, some species could have suffered translocation events which have led to the gain of an additional 5S rDNA locus. The translocation to new chromosomal loci has traditionally been explained by unequal crossing-over between non-homologous chromosomes or by a transposon-mediated mechanism . The signals from these six Haemulidae species were located either in telomeric position (in all species) of the q arm or near the centromere (in 4 out of the 6 species). The first location could be the pleisiomorphic condition in the Haemulidae family, which has been lost in P. mediterraneus and has retained the more conservative pattern of the 5S rDNA localization.
The ITS-1 size and GC content were very close to the average values for the Osteichthyes group (635.1 bp and 68.0% respectively) . Two species of the Haemulidae fish family showed ITS-1 sequences particularly similar to that of P. mediterraneus. The ITS-1 region has secondary structures necessary for the rRNA maturation process, which give this region an intermediate rate of evolution . These features make the ITS-1 region suitable for phylogenetic studies among closely-related groups . Considering that P. mediterraneus showed an ITS-1 region more similar to that of Parapristipoma trilineatum than to that of the same genus, Plectorhinchus cinctus, a phylogenetic revision of the Haemulidae family should be addressed.
At the cytogenetic level, the 18S rRNA gene probe did not conserve the proposed plesiomorphic localization for fish species, i.e. one chromosome pair with signals in the telomeric position nearest to the centromere. This affirmation has been made based on Ag-NOR studies , but other studies using FISH techniques with major ribosomal probes have also confirmed it [10, 11, 57, 62–65]. To the contrary, P. mediterraneus presented a derived (apomorphic) pattern in which the major ribosomal probe hybridizes in the subcentromeric region. Moreover, this localization is common in all Haemulidae species studied so far using both Ag-NOR staining and rDNA-FISH methods [54–55, and references therein]. Therefore, the pattern shown by P. mediterraneus is the plesiomorphic condition within the Haemulidae family. The non-syntenic arrangement of both ribosomal clusters, as we have found in P. mediterraneus, is the most commonly-observed situation in fishes . The transcription by different RNA polymerases has traditionally been given as the explanation for this lack of co-localization . Moreover, the separated localization could prevent undesirable disruptive translocations of the 5S rDNA into the major ribosomal cluster .
The U2 snRNA gene has been found linked to another snRNA: the U5 snRNA gene. This seems to be the common situation among the fishes so far studied with this gene family, since the same linkage has been seen in two species from the Moronidae family , four from the Engraulidae family  and in Oreochromis niloticus, and the U1, U2 and U5 snRNA linkage has also been observed within the Soleidae family [9, 67]. In Drosophila species, different types of linkages have been observed, such as U1-U2, U2-U5 and U4-U5 linkages . The linkage of different members of the snRNA gene family could serve as a means of consolidating phylogenetic lineages; indeed  were able to prove that U1-U2 linkage was present in the Soleidae family but not in the Scophthalmidae and Pleuronectidae families.
Moreover, by aligning both 5’ and 3’ ends of the two U2-U5 spacers with those from other fish species, it was possible to find four conserved regions which may be related with the 3’ box and the proximal sequence element (PSE) of the two genes. The 3’ box is necessary for the transcription termination and for the correct processing of the nascent mRNA , and is positioned at between 9 and 19 nucleotides (nt) downstream from the coding region . The two 3’ boxes found in P. mediterraneus are a little nearer to the coding region. The PSE is essential for the initiation of transcription and is found between 40 and 60 nt upstream from the transcription starting point [69, 70]; this is where the two PSE found in P. mediterraneus are situated. A third element, the distal sequence element (DSE), is normally located between 200 and 250 nt upstream from the PSE . In the case of P. mediterraneus, several putative octamer motifs similar to that proposed for all vertebrate DSE  have been found (data not shown), but none of them are located at the cited nt range upstream from the PSE. The presence of all these regulator elements makes the U2 and U5 snDNA transcriptionally active genes.
The chromosome localization of the U2 snRNA gene has scarcely been studied in the fish group. In two species from the Moronidae family this probe hybridizes in the telomeric region of one acrocentric pair , and in four species from the Batrachoididae family it hybridizes in three different manners: dispersed, located in the subtelomeric position of one acrocentric pair, and both dispersed and located . None of these cases share the same hybridization pattern as that observed in P. mediterraneus, and this makes the U2 snRNA gene of particular interest for cytotaxonomic purposes. To date only one case has been reported in fishes of co-localization between snRNA and ribosomal genes by means of the FISH technique, as is found in Thalassophryne maculosa (Batrachoididae family) . Moreover, in Solea senegalensis, the 5S rDNA and the U1, U2 and U5 snDNA were found closely linked . However, the non-syntenic organization of snRNA and ribosomal genes also seems to be the most common arrangement in fishes studied so far.
The spacer 2 of the U2-U5 cluster showed an exceptional low value of nucleotide variability with respect to the spacer 1 and the U2-U5 coding regions, and this peculiarity may be due to the small size of the spacer 2 region (168–178 bp). This size could be close to the minimum necessary for maintaining the U2 array and so only small variations can be afforded, in a similar way as had previously been proposed for short 5S arrays . Similarly, the NTS type γ also presents low variability because the external promoters are included in a short NTS of 72 bp, and only slight variations should be tolerated. However, the same maintenance does not appear to occur in the NTS type δ, since its size is 61 bp long and the nucleotide variability is rather higher than the other NTSs. As has been mentioned before, the 5S rDNA type δ is probably a pseudogene, and neither the coding region nor the external promoters are bearing selective pressures that are homogenizing the array; therefore it is free to overcome mutations, thus increasing the nucleotide variability. The ribosomal genes (18S, 5.8S and 5S) showed lower levels of variability than the snRNA genes (except in the type δ of the 5S), thus indicating that the homogenizing forces of concerted evolution are more efficient in the ribosomal loci. Despite the nucleotide variability observed and the presence of various indels in the U2-U5 cluster, the different clones maintain a considerable homogeneity in their sequences, and no variants and pseudogenes were observed. Therefore, both major ribosomal and snRNA genes evolve according to the concerted model.
The karyotype obtained in P. mediterraneus shares the same karyotype formula (2n = 48; FN = 48) of most of the species of the Haemulidae family [54, and references therein]. Traditionally, it has been postulated that this karyotype formula is the ancestral (plesiomorphic) formula for the Teleostei subclass ; nevertheless,  suggested that the 2n = 48 acrocentric chromosome formula should have arisen later, during teleost spreading, and probably was the plesiomorphic karyotype for the Clupeomorpha and Euteleostei superorders. From the FISH analysis three chromosomal markers were obtained for the P. mediterraneus chromosome complement, due to the non-syntenic organization of the three probes used. Obtaining chromosomal markers is particularly important for distinguishing unequivocally chromosomal pairs from karyotypes composed of very small chromosomes and with similar sizes between adjacent pairs.